Mergers during the first and second phase of globalization: Success, insider trading, and the role of regulation

Mergers During the First and Second Phase of Globalization:

Success, Insider Trading, and the Role of Regulation

Inaugural Thesis

To Obtain the Doctoral Degree of the Faculty of Economics Eberhard-Karls-Universität Tübingen

Germany

Submitted by

Gerhard Kling Born in Munich

2004

Dean: Professor Dr. rer. pol. Renate Hecker First Referee: Professor Dr. pub. oec. Jörg Baten Second Referee: Professor Dr. rer. pol. Werner Neus Day of the Disputation: 28^th May 2004

Gerhard Kling

Mergers During the First and Second Phase of Globalization:

Success, Insider Trading, and the Role of Regulation

“I find it difficult to think of economists and economic historians as separate animals. Their interests are fundamentally the same. The job of the economist is to explain how the economy works; the job of the economic historian is to explain how it worked in the past.”

A. K. Cairncross

Acknowledgements

First of all, I take the opportunity to thank my advisor Jörg Baten for his steady support during `periods of suffering and doom´ that are typical for a dissertation project. He also encouraged me to attend conferences and present my research to an international audience.

Even more important, he gave me the freedom to develop my own thoughts, make my own mistakes, and gain experience.

My thesis is based on five working papers – but they all contribute to one single story that highlights the success of mergers in Germany in different periods of time. Two of these working papers were presented at international conferences; hence, I want to thank for the many comments I received. My paper entitled `The impact of merger announcements on stock prices: The rejection of the merger paradox for German companies´ was presented at the annual meeting of the Business and Economic Historical Society in Memphis 2003. Lynne Pierson Doti, as my discussant and chair of the session, and Jari Eloranta gave me some very useful comments. This paper was also presented at the 5<sup>th</sup> European Historical Economics Society Conference in Madrid 2003. Thus, I want to thank Stefano Battilossi, Albrecht Ritschl, and Joachim Voth for the challenging debates. In the general discussion, Stephen Broadberry asked whether mergers were successful in the long-run; now, I am able to respond, and my fifth chapter is dedicated to this task. My paper on the `Disclosure of mergers without regulatory restrictions: Comparing insider-trading in the year 1908 and 2000 in Germany´ was invited for presentation at the NBER conference on Developing and Sustaining Financial Markets, 1820-2000, which took place 2003 in Boston. I thank Marc Weidenmier for his excellent comments as well as Lance Davis, Larry Neal, and Eugene White who organized the conference and contributed remarkably to the high quality of discussions. I had also the chance to participate in the research seminar for PhD students organized by the German Finance Association (DGF) (2003 in Mainz) and in the Third Summer School in Institutions, Economics, and History: Credit Networks and Economic Development (2003 in Venice). On 26^th January 2004, I presented the joint paper that I wrote together with Markus Baltzer entitled `Resiliency of the pre-World War I German stock exchange: Evidence from a panel vector autoregression´ at the Workshop in Economic History at the Humboldt University in Berlin. Furthermore, I took the chance to present my research several times at the Economic Workshop at the University of Tübingen. Consequently, I received many comments and suggestions to further improve my work. For their active role in the discussions, Werner Neus and Joachim Grammig deserve special thank.

The scholarship from the German Research Foundation (DFG) supported my research considerably and mitigated the financial burden imposed by attending international conferences. I also want to thank the German Finance Association (DGF), the department of economic history of the Humboldt-Universität, and the National Bureau of Economic Research (NBER) for their financial support.

Furthermore, I thank the program committee that selected my paper on the `Disclosure of mergers´ for the annual meeting of the Economic History Society. Due to the `self- sacrificing´ work of my co-author Markus Baltzer our paper entitled `Resiliency of the pre- World War I German stock exchange: Evidence from a panel vector autoregression´ attracts international interest; accordingly, we will present our paper at the European Social Science History Conference 2004 in Berlin. Even more noteworthy, this paper was also accepted for presentation at the 5^th World Congress of Cliometrics that will take place in Venice 2004.

The encouraging atmosphere in the graduate school and in our research group in economic history facilitated my work considerably. Henceforth, I thank all members of the graduate school and our research group for their friendship and steady support. I should emphasize, especially, the excellent comments due to Alexander Moradi and Aravinda Meera’s advice with regard to language, spelling, style, and other `horrible´ things.

Preface

Before you start reading, I want to highlight some basic convictions I have with regard to language, cited literature, and presenting problems inherent with data sets and empirical methods. I strongly believe in these convictions and I am convinced that they contribute to the quality of research.

The primary goal of scientific writing is to maintain clarity. Studying the books written by Day (1995, 1998) and Strunk and White (1999) helped to come closer to this goal – but it stays a difficult challenge. To avoid too wordy expressions, Day (1995, 1998) preferred to use active voice; thereby, `I´ and `We´ are interchangeably – but uncommon for German ears. By the way, I also learnt how to use semicolons in English – albeit I have still no idea how to use semicolons in German. Nevertheless, I hope that my language is simple enough to convey the sophisticated content.

Unfortunately, economic history is an extremely broad area of research that produces tons of literature every year; hence, I restricted myself to essential contributions published in refereed journals or outstanding edited volumes. Even worse, my research also includes topics in finance and econometrics which increases the related literature further. Accordingly, I cite only important sources that contribute to my research considerably. Generally, I focus on the

`working paper style´ which avoids too lengthy reviews of literature. In contrast, my own data sets, methods, and results are of primary interest. Nevertheless, it is valuable to stress to what extent my research contributes to the existing strands of the literature. Of course, the typical literature in economic history usually contains hundreds of references; however the following quote that is due to William C. Roberts expresses my conviction best. “Manuscripts containing innumerable references are more likely a sign of insecurity than a mark of scholarship”.

Of course, empirical researchers often want to produce significant results, and only these results are usually published; however, having no results also contributes to science.

Henceforth, I strongly believe that highlighting the limitations of data sets and methods is a crucial part of empirical research. For instance, I am very proud that I fail to detect a long- term impact of mergers on share prices and dividends. Thomas A. Edison put it in the following manner: “Results! Why, man, I have gotten a lot of results. I know several thousand things that won’t work.”

Often used symbols and abbreviations Pt Stock price

Eit Error term

σe2 Variance of the error term e which is equal to the residual of the CMR model µi Mean return of stock I

Rit Observed daily returns

A Matrix that contains for all stocks i all observed daily returns for the whole

estimation period

L Length of the estimation period εit* Abnormal return

τm or τn Indicate a specific point in time; thereby, m ≤ n

(

Cˆ

)

∗

εt Portfolio weighted abnormal return

(

C τ ;τ

)

Nit Nominal capital

bt-1 Estimated parameter vector for t; thereby, one uses only the information available at t-1

mit Executed merger

∆zit Vector that contains the first difference in share prices and in dividends inft Inflation rate at time t

Σj Coefficient matrix for lag j of the VAR in reduced form

gt Vector that contains unexpected shocks in inflation and growth rates CMR Constant mean return model

MM Market model

CUSUM Cumulated sum of residuals

GARCH Generalized autoregressive conditional heteroscedasticity model ARCH Autoregressive conditional heteroscedasticity model

ARIMA Autoregressive integrated moving average model VAR Vector autoregression

Contents

1. Introduction

1.1 The aims of my dissertation project 1

1.1.1 Measuring the success of mergers 1

1.1.2 Who gains from mergers? 2

1.1.3 Methodological issues and alternative approaches 3

1.1.4 The long-term impact of mergers 3

1.2 Why should we care about historical evidence? 4 1.3 Is this applied econometrics, finance or economic history? 5

1.4 The structure of my dissertation 5

2. The impact of merger announcements on stock prices

2.1 Extended abstract 7

2.2 Introduction 7

2.2.1 Former studies on the merger paradox 7 2.2.2 What makes my analysis special? 10 2.2.3 The event-study method: A brief review 10

2.2.4 The structure of this chapter 11

2.3 The method of sampling 11

2.3.1 What can be regarded as a merger? 11 2.3.2 What can we learn from this case study? 14

2.3.3 Drawing a sample 14

2.3.4 Why should I choose the year 1908 as investigation period? 16

2.3.5 Descriptive analysis 18

2.3.6 The role of cartels and syndicates 20 2.4 The theoretical background of event-studies 24

2.4.1 Random walk hypothesis 24

2.4.2 Merger announcements – events with great influence 24 2.4.3 The constant-mean-return (CMR) model 26 2.4.4 Abnormal returns and their statistical properties 27 2.4.5 Aggregation of abnormal returns over time 28 2.4.6 Cumulating abnormal returns over time and over cross sectional units 29

2.5 Empirical results of the event-study 30 2.5.1 Results for the estimation period 30 2.5.2 The justification for the choice of the event period 32 2.5.3 Abnormal returns and cumulated abnormal returns of the whole sample 35

2.5.4 Division into two subgroups 38

2.6 Cross sectional model 43

2.7 Conclusion 53

3. Disclosure of mergers without regulatory restrictions: Who gains from mergers?

3.1 Extended abstract 56

3.2 Introduction 56

3.3 Historical Background – Insider Regulation in 1908 and 2000 59 3.3.1 Legal Framework in the year 1908 in Germany 59 3.3.2 Ad-hoc-publication and insider trading in the year 2000 60

3.4 Method of Sampling 61

3.4.1 Determination of the event day 61

3.4.2 Should one include unsuccessful mergers? 62 3.4.3 Determination of the event and estimation period 65

3.5 Event-study analysis 65

3.5.1 `Recalling´ the basic concept 65

3.5.2 Results of the estimation period in the year 1999 67 3.5.3 Abnormal returns and cumulated abnormal returns in the year 2000 69 3.5.4 Does the market know if a merger fails to overcome the hurdles? 69 3.5.5 The way of disclosure in the year 1908 73 3.5.6 Was the regulation of insider trading successful during the last 92 years? 77

3.6 Cross-sectional analysis 79

3.6.1 Are undisclosed mergers in 1908 more successful? 79 3.6.2 If managers expect profitable mergers, they will hide information 84

3.7 Conclusion 87

4. The limitation of event-study analysis: Problems and alternative methods

4.1 Extended abstract 89

4.2 Introduction 89

4.3 General problems using the event-study approach 90 4.3.1 How important is the length L of the estimation window? 90 4.3.2 Is it possible to detect abnormal returns in time periods without an event? 91 4.3.3 Is the constant-mean-return model inferior in comparison to the market model 94 4.3.4 Do dividend payments affect the results? 96 4.3.5 Shareholder value orientation versus redistribution theory 97 4.4 Explaining the adaptation process of stock prices: A traditional view 97 4.4.1 Extension of the cross-sectional model 97 4.4.2 How can the time path of abnormal returns be modeled? 100

4.4.3 The recursive CUSUM approach 101

4.4.4 The OLS based CUSUM test 105

4.4.5 The failure of the traditional approach 105 4.5 Explaining the adaptation process of stock prices: A time series approach 108 4.5.1 Intervention models with transitory shocks 108

4.5.2 Transfer function models 111

4.5.3 Different ways of disclosure and the `empirical announcement day´ 115 4.5.4 Does speculation affect the `empirical announcement day´? 116 4.5.5 The limitations of the transfer function analysis 116 4.6 Event-induced uncertainty in daily abnormal returns 118 4.6.1 Developing a basic panel based GARCH approach for abnormal returns 118 4.6.2 Specifying the correct GARCH(p, q) model 119 4.6.3 The GARCH(1,1) model with and without stock specific effects 121 4.7 Non-synchronous trading and information from stock price jumps 124 4.7.1 How important is non-synchronous trading? 124 4.7.2 What can we learn from the trading patterns? 128

4.8 Conclusion 131

5. The long-term impact of mergers and the role of macroeconomic shocks

5.1 Extended abstract 133

5.2 Introduction 133

5.2.1 The long-term impact of mergers 133 5.2.2 Share prices and macroeconomic shocks 134 5.2.3 The classical gold standard and macroeconomic stability 136

5.2.4 The emergence of merger waves 137

5.2.5 Structure of this chapter 137

5.3 Method of sampling 138

5.3.1 How to construct a representative sample? 138 5.3.2 Inflation rates and economic growth 142 5.3.3 Testing for unit-roots in share prices and dividends before and after deflating 144 5.3.4 Missing values and the Holt-Winter filter 144 5.3.5 Annual data on mergers: Discussing the pros and cons 145 5.3.6 Share prices, dividends, and nominal capital in different industries 147 5.3.7 How important is the survivorship bias? 150 5.4 What drives merger during the first phase of globalization in Germany? 153 5.4.1 How should I model the driving forces for mergers? 153 5.4.2 Merger activity during the investigation period 1870 to 1913 153 5.4.3 Model selection of a dynamic panel probit model with random effects 156 5.4.4 Predicting mergers during the period 1870 to 1913 159 5.5 The anticipation of macroeconomic variables 160 5.5.1 Why is it essential to talk about anticipation of macroeconomic conditions? 160 5.5.2 Can one anticipate future inflation and growth rates? 161 5.6 Panel vector autoregression with macro and microshocks 163 5.6.1 `Traditional´ VAR model with prices, dividends, and macroeconomic factors 163 5.6.2 Panel VAR framework with share prices, dividends, and exogenous shocks 163

5.6.3 Determining the lag length p 168

5.6.4 Why is the change in nominal capital not considered in my panel VAR? 168 5.7 Empirical findings: Macro and micro-level shocks 170 5.7.1 Outcomes of the VAR in reduced form – the importance of macro shocks 170 5.7.2 Outcomes of the VAR in reduced form – mergers or forecasting errors 172 5.7.3 Impulse response functions 174 5.7.4 Bootstrapping method to derive CI for impulse response functions 178 5.8 The long-term relation between share prices and dividends 178

5.8.1 Introduction 178

5.8.2 Cointegration relation between share prices and dividends 179 5.8.2.1 Transformation of time series and hidden cointegration 179 5.8.2.2 Individual tests for cointegration 183 5.8.2.3 Testing for cointegration in panels 186 5.8.2.4 Estimating the cointegration vector between prices and dividends 190

5.8.2.5 Interrelation between merger activity and real valuation of stocks 193 5.8.3 Decomposing the transitory and permanent components of time series 197

5.8.3.1 The econometric procedure 197

5.8.3.2 Calculation of canonical correlations 198 5.8.3.3 Components of the `Berliner Handelsgesellschaft´ 199 5.8.3.4 Transitory perturbations and the merger activity 202 5.8.3.4.1 The real valuation of the whole market and the merger wave 202 5.8.3.4.2 The real valuation in different industries and the merger activity 202

5.9 The role of the exchange law 1896 207

5.10 How important are mergers for the expansion of enterprises? 208

5.11 Conclusion 211

6. Concluding remarks

6.1 The success of mergers 213

6.1.1 Rejecting the merger paradox for the pre-World-War I period 213 6.1.2 The emergence of the merger paradox in the year 2000 213 6.2 Which shareholders gain from mergers? 215

6.2.1 Insider trading in the year 1908 215

6.2.2 Irrational speculation in the year 2000 216 6.3 Interpreting the high informationally efficiency in 1908 216

6.4 The long-term impact of mergers 216

6.5 Merger waves and periods of real over- or undervaluation 217 6.6 What makes mergers different: Comparing both phases of globalization 217 6.7 The inflation illusion hypotheses and the pre-World-War I period 218

6.8 Was the effort worth it? 220

7. Mathematical and statistical appendix

7.1 Mathematical appendix 223

7.1.1 Technical note for equation (2.6) 223

7.1.2 Proof of proposition 1 223

7.1.3 Proof of proposition 2 226

7.1.4 Proof of Proposition 3 226

7.1.5 Proof of Proposition 4 227

7.2 Statistical appendix 227 7.2.1 Bootstrapping to construct confidence intervals for impulse response functions 227 7.2.2 Decomposing time series into transitory and permanent components 232

8. References

8.1 Cited literature 234

8.2 Data sources and law texts 247

8.2.1 Law text 247

8.2.2 Data sources 247

1. Introduction

1.1 The aims of my dissertation project 1.1.1 Measuring the success of mergers

The merger wave that took place during the first phase of globalization, which lasted from 1895 to 1914,¹ changed the industrial structure in Europe and the U.S. remarkably. Therefore, it is of great importance to assess whether mergers were successful during this period.

Noteworthy, studies that evaluate the success of mergers during the first phase of globalization are still lacking for Germany. One may argue that this statement is false and could refer to Huerkamp (1979). However, she defined success – by construction of her sample – as the ability of a firm to stay among the 100 largest companies.² Hence, shareholder value destroying mergers driven by `empire building´ that increase the firm size are seen as successful investments. Consequently, I totally disagree with her view. In contrast, I try to quantify the market response due to mergers and, hence, focus on the change in shareholder value. This imagination is in line with studies on the success of mergers for the United States and Great Britain. Generally, for the German case, economic historians concentrated on debates about the interrelation between the expansion of large scale enterprises, external growth, and mergers.<sup>3</sup> Maintaining size and survivorship were seen as major factors of success. But also `traditional´ cross-country studies showed that the large German enterprise was a main guarantee for superior economic development in the pre-World-War I period.⁴ After reviewing new statistical material, however, the picture has to be corrected. A recent empirical cross-country study on that issue was written by Kinghorn and Nye (1996). They found evidence that German firms and production facilities were smaller compared to U.S. or French companies.⁵ In addition, the concentration process was less developed in Germany.

Besides these astonishing results, additional doubts emerge regarding the alleged success of large firms. In several empirical studies, Baten (2001 a, b, c) showed that small firms

1 A subsequent session will discuss the so called `first phase of globalization´ in greater detail.

2 Note that she used the data set collected by Kocka and Siegrist (1979); thereby, the 100 largest firms in 1907 were included – but their merger activity from 1887 to 1907 was studied. Hence, only companies that stayed among the 100 largest firms during this period were considered. Tilly (1982) argued that success can be defined in their study as the ability to stay among the 100 largest companies over the whole period.

3 Tilly (1982, 1986), Gerschenkron (1962), Huerkamp (1979), and Feldenkirchen (1988) discussed the concentration process in different lines of business.

4 I refer to the most famous contribution made by Chandler (1990). For late Victorian Britain, Elbaum and Lazonick (1986) argued in favor for large enterprises that had the ability to adopt the modern form of corporation; thereby, they referred to the German and U.S. example.

5 Kinghorn and Nye (1996) calculated the number of workers in different industries for Germany, U.S., and France; thereby, German companies are relatively small. Especially, in the iron and steel industry, U.S.

companies employed on average about two times more workers than German counterparts.

exhibited a larger total factor productivity. Moreover, he provided evidence that contradicted the usual opinion suggesting a steady increase in firm size between 1895 and 1912. In contrast, he found that the median of firm size stayed unchanged over time.⁶

When one turns to studies for the United States of America or Great Britain the scope regarding mergers is totally different compared to the `traditional´ research conducted in Germany. For instance, Leeth and Borg (1994, 2000) who covered the years 1905 to 1930 measured the economic impact of mergers by applying event-study methodology. In their study, successful mergers should yield an upsurge in market value.

Accordingly, my first aim is to assess the success of mergers based on the market response caused by merger announcements; thereby, a higher market value is the recipe for success and not firm size. Encouraged by the results of Baten (2001 a, b, c), I also collect data on mergers among smaller companies, which was, thus far, not done. Of course, my research contributes to close the data gap for Germany that is due to the absence of sources like Nelson (1957) and Eis (1971) who systematically collected data on mergers among U.S. companies.

1.1.2 Who gains from mergers?

If I, indeed, detected an increase in market values stemming from a merger announcement, another question would arise. Which type of shareholder gains from higher market values?

Focusing on two types, namely insiders and outsiders, my aim is to answer this question;

thereby, the so called run-ups prior to merger announcements serve as a measure for insider gains. Run-ups are changes in stock prices triggered by an impending merger announcement.

As long as the merger is not yet public information, significant changes before the public release serve as a hint for insider-trading. If a market participant has only access to public sources like the official newspaper announcement, this participant belongs to the group of outsiders. In contrast, insiders possess private information; hence, they already know that a firm will announce publicly that they engage in merger activities. This superior knowledge leads to trading activities of insiders before the public announcement. Through this insider trading the private information is conveyed; thus, the market price is significantly influenced.

Keown and Pinkerton (1981) used this measure to uncover insider activities around revealed mergers occurring in the years 1975-1978. Banerjee and Eckhard (2001) provided evidence for insider-trading in the year 1896-1903 known as the first merger wave. Both studies concentrate on the U.S. case.

6 Due to data availability, his research is restricted to `Baden´; however, his empirical evidence can also be interpreted as a general statement regarding firm size. An exception, however, are some regions in which large scale enterprises (iron and steel, mining etc.) predominated the industrial structure.

Lacking regulatory restrictions are responsible for the appearance of two different forms of disclosure in the pre-World War I period in Germany. Some firms announce mergers after these mergers have already been executed, and others declare their desire to merge before the transfer of assets. Thus, one consideration is to assess whether the way of disclosure influences the gains respectively losses for insiders and outsiders. By comparing the pre-Word-War I period with mergers that took place in the year 2000 in Germany, I try to shed some light on the impact of regulations on insider activities and the ability of legislative restrictions to protect outsiders from insider trading.

1.1.3 Methodological issues and alternative approaches

Besides discussing economic issues like the change of shareholder value caused by mergers, I also try to thoroughly highlight methodological concerns. Consequently, I extend my former event-studies and conduct consistency checks to prove whether market frictions like non- synchronous trading affect my results. Caused by many restrictive assumptions inherent with event-study methods, I propose an alternative transfer function model. This time series approach has the capability to detect an `empirical announcement day´; hence, this model serves as an alternative to identify run-ups. Furthermore, the microstructure of the Berlin stock exchange around 1900 can be explored by determining periods during which no trade is executed. These non-synchronous trading causes frictions that may influence my former results. Consequently, my thesis should also contribute to solve methodological issues.

1.1.4 The long-term impact of mergers

Using event-studies, I concentrate, thus far, on short-term market reactions caused by mergers. My additional concern is to shed some light on the long-term impact of mergers;

thereby, an event-study approach must be replaced by more sophisticated methods. These superior models belong to the group of vector autoregressions (VAR). Besides focusing on mergers and, thus, micro-level shocks, I regard macroeconomic fluctuations as additional source of uncertainties. My panel VAR identifies the dynamics in share prices, dividends, and nominal capital caused by different kinds of shocks. In contrast to my short-term analyses, my long-run study covers the period from 1870 to 1913; thereby, I collected annual data. Changes in the regulatory environment at the beginning and in the middle of this period – especially the establishment of the new exchange law in 1896 – make the investigation promising from an institutional point of view.

1.2 Why should one care about historical evidence?

At conferences it is just a matter of time that the question arises whether we should care about mergers that occurred approximately 100 years ago. Or to put it differently, how can historical evidence improve our understanding regarding current mergers that typically take place during merger waves. I think the sense for looking at historical data is threefold. First, as mentioned, mergers normally occur in waves that affect single industries or the economy as a whole. The first merger wave can be located around 1900 and was a worldwide phenomenon.

Collecting data on mergers during the first merger mania may help to clarify the emergence of the last merger wave that occurred shortly before the `bubble´ burst in the year 2000.

Second, globalization is not a new experience and makes a comparison between the first and second phase of globalization promising. Sachs and Warner (1996)⁷ pointed out that already at the end of the nineteenth century a liberal international economic order allowed a global capitalism. This phase of liberalization is to some extent comparable to the current advances of a global economy. When one considers labor mobility and migration, the late nineteenth century was much more globalized than the late twentieth century.⁸ The urgent question is whether the globalization about 100 years ago is similar to the integration process nowadays.⁹ Besides other discrepancies, the emergence of multinational enterprises and the importance of intra-industry trade is a historically new experience as mentioned by Bordo et al. (1999). Caused by companies that sliced up their value chain by different sorts of foreign direct investment,¹⁰ intra-industry trade becomes predominant in the trade between developed and developing countries. In contrast, in the late nineteenth century, the periphery mainly exported primary goods to developed countries. Assessing the degree of financial integration is also important for my research. However, the outcomes are ambiguous. Vásquez (2000) stressed¹¹ that the world is nowadays less financially integrated compared to the situation 100 years ago if one uses net capital flows as measure. In contrast, Bordo et al. (1999) and also the experiences from the Asian crises in 1997 showed that capital flows nowadays react much faster. This is due to the increasing importance of portfolio investments.

Consequently, the debate on the similarities and differences between both phases of globalization is still enduring; hence, my research will contribute some evidence on merger activities to this overall discussion.

7 Williamson (1996) and Tilly (1999) besides others coined the term `first phase of globalization´; however, the exact location of this period exhibiting an increase in international integration is disputable.

8 Hatton and Williamson (1994, 1998) discussed the mass migration from Europe.

9 A current study on this subject by Bordo et al. (1999) discussed some issues and concluded that the current globalization of commodity and financial markets is historically unprecedented.

10 This may also include cross-border mergers.

11 This argument is based on figures obtained from the World Economic Outlook (1997).

Third, I can observe the `natural´ behavior of firms acting in an environment without regulatory restrictions. This historical experiment enables to quantify institutional changes and to derive policy recommendations whether state interventions are welfare creating – or not. Empirical investigations can also figure out why some firms behave nicely while others cheat. The third chapter, especially, deals with this advantage of analyzing historical periods.

1.3 Is this applied econometrics, finance or economic history?

Generally, the success of mergers and acquisitions is a central topic in corporate finance;

however, I work with historical data, which points clearly into the direction of economic history. In addition, the econometric content of my thesis goes obviously far beyond

`standard´ contributions in economic history. Therefore, someone may ask which label my research should have. Out of my point of view, my thesis combines different areas of specialization and may be regarded as a `jagged´ alliance among economic history, corporate finance, and applied econometrics. Nevertheless, my research can be characterized as quantitative economic history and cliometrics respectively. Hence, someone who expects only case studies and qualitative discussions might be disappointed. In contrast, my thesis is oriented toward empirical research and favors a more or less rigorous treatment of the applied econometric techniques.

1.4 The structure of my dissertation

My thesis can be split into two major parts; thereby, the second, third, and fourth chapter discuss the short-term economic impact of merger announcements. In the fifth chapter, I turn to the long-lasting influence of mergers on company characteristics like share prices, dividends, and the nominal capital. More precisely, the second chapter tackles the challenge to measure the market response triggered by merger announcement by relying on daily returns. Using daily returns is new for the pre-World-War I period. Thereafter, chapter three quantifies the scale of insider trading and the role of regulation. Comparing different regulatory frameworks established over time in Germany, I can assess the effectiveness of regulations. Observing unrestricted firm behavior allows to figure out whether a mechanism of self-regulation may work or the state should intervene. Although event-studies are very useful and widely applied to detect short-term market reactions, there may be alternative approaches. Hence, chapter four introduces modified transfer function models that overcomes usual pitfalls of event-studies. Besides the short-term market reactions, the long-term impact of mergers attracts my interest. It turns out that this task is highly challenging from an

econometric point of view; thus, it deserves a thorough and rigorous analysis in chapter five.

As a `waste product´ of the fifth chapter, several models are developed that clarify the expansion of enterprises and the decision to undertake a merger. A broad discussion of my results and an outline of future research topics conclude my thesis.

2. The impact of merger announcements on stock prices

2.1 Extended abstract

The scope of this chapter is to capture the market response triggered by merger announcements. Hence, the stock market decides whether a merger can be regarded as success. The construction of my short-term investigation enables to evaluate the performance of acquiring and target firms. Consequently, I can answer the question whether the merger paradox is observable in the pre-World-War I period in Germany. An event-study method applied to daily returns rejects the merger paradox based on my data. In addition, the adaptation process of stock prices according to newly available information is finished within a few days around the event day. Correspondingly, the exchange seems to be highly informationally efficient. To detect what affects the success of mergers, I use cumulated abnormal returns as dependent variable in a cross-sectional study. Controlling for direct and indirect effects, I construct a simultaneous equation approach. I uncover that banks exhibit remarkably positive abnormal returns in comparison to other lines of business.

2.2 Introduction

2.2.1 Former studies on the merger paradox

Thus far, a study that uses market valuations to quantify the success of mergers is still missing for the pre-World-War I period in Germany. Encouraged by the excellent availability of data sources¹² on executed mergers during this period in the United States, several event-studies for the U.S. were published. The analysis conducted by Leeth and Borg (1994, 2000) who covered the period from 1905 to 1930 attracted international interest. Furthermore, for the so called second merger wave that took place in the 1920s, studies did already exist.¹³ Nevertheless, the majority of the literature in empirical finance started to measure the performance of acquiring and target firms around public releases of mergers occurring in the 1960s. Generally, evidence for the 1980s and later periods suggested that only the shareholders of target firms gain from takeovers and the share prices of acquirers are nearly unaffected.¹⁴ Moreover, some empirical studies, for instance Travlos und Papaioannou (1991), found even negative abnormal returns of the acquiring firms. Why should firms

12 As mentioned in chapter one, Eis (1971) and Nelson (1959) are by far the most cited sources for the United States.

13 See Borg et al. (1989).

14 Jarrell and Poulsen (1989) found positive abnormal returns for target firms – but negative or insignificant abnormal returns for acquiring firms.

initiate mergers if this means a loss for their shareholders? This finding is often called the

`merger paradox´; however, there are several plausible solutions to clarify this puzzle.

One widespread idea is that institutions like monopoly commissions or more general the legal framework play a crucial role whether mergers create shareholder value. Jarrell and Poulsen (1989) pointed out that the lacking restrictions in the 1960s and 1970s compared to the 1980s are mainly responsible that the merger paradox can be confirmed for the latter period. However, compared to the pre-World-War I period, the 1960s or 1970s are highly

`over- regulated´.¹⁵

The introduction of junk bonds in the 1980s facilitates the access to capital and, thus, is also seen as an essential factor to provoke a negative market reaction after merger announcements. If it is too simple to raise up money to finance an external expansion, inefficient mergers are more likely to be executed. One should take into consideration that larger companies have advantages in financing a merger. Correspondingly, Jarrell and Poulsen (1989) tried to control for this issue by considering the ratio of the firm size between the acquirer and the target firm. Besides the access to capital, they claimed that the larger the acquirer relative to the target the less important is the acquisition and, hence, the less likely is a negative market response.

The moral hazard problem inherent with the separation of ownership and control is regarded by many authors as additional source for the merger paradox. Thus, Jarrell and Poulsen (1989) argued that a manager favors even a shareholder value destroying merger to make the company larger. A larger company weakens the possibility for shareholders to control effectively the management. Shleifer and Vishny (1988) underlined that managers enjoy to increase their influence by `empire building´ that describes external growth for the sake to becoming larger – but not to increasing shareholder value. Based on these statements, there is also the imagination of the `market for corporate control´ that is due to Manne (1965).

The prerequisite for this incentive mechanism is a highly positive correlation between the market valuation of a company and the quality of its management. If a management is inefficient, the risk of a tender offer¹⁶ or a takeover bid will increase caused by the lower market value of the badly managed company. Accordingly, this external threat works as an incentive for the management to focus on the maximization of shareholder value.

15 Borg et al. (1989) argued in a similar manner and favored their period 1919 to 1930 because it is less regulated. Nevertheless, my investigation period is even more liberal concerning mergers, cartels, and collusive behavior in general.

16 The SEC defines a tender offer as “(...) broad solicitation by a company or a third party to purchase a substantial percentage of a company’s shares or units for a limited period of time.”

How can my historical data set contribute to the understanding of the merger paradox?

Compared to the 1960s and 1970s, the scale of regulation regarding horizontal mergers and collusive behavior was extremely low respectively did practically not exist in Germany during the pre-World-War I period. A monopoly commission, for instance, was not established, and other legal thresholds or local authorities were seldom an obstacle for mergers. More specifically, cartels and syndicates were part of the scene of the German industry,¹⁷ albeit public opinion did not favor collusion.¹⁸ In addition, political debates regarding cartels were quite common and intensified by events like the `coal need´ in 1900/1901 that led to a pronounced increase in coal prices.<sup>19</sup> Correspondingly, the first merger wave that occurred from 1898 to 1904 – based on the investigation by Banerjee and Eckhard (2001) – can be characterized as `mergers to monopoly´ as described by Stigler (1950).

Obviously, the separation of ownership and control was not predominant in the pre- World-War I period and, especially not, in small or medium sized companies. The manager of a smaller company was very often also the largest or at least an important shareholder.

Furthermore, members of the advisory board had a considerable stake in the company; hence, typical free rider problems did not prevent effective control of the management. It is also noteworthy that the first concerns about incentive and control problems inherent with the separation of ownership and control were discussed by Berle and Means (1932) if one accepts the view of Scherer (1988). However, Pitelis (2004) pointed out that the incentive problems were already detected by Knight (1921) or are due to the “founding father of economics”

Adam Smith (1776).

According to these highlighted discrepancies between the pre-World-War I period and the second phase of globalization, which starts after the second World War, one should expect that the merger paradox did not exist in the former period. This empirical finding would be also in line with former studies for the U.S. industry for the first merger wave²⁰ and the second merger wave in the 1920s. Thus, one may argue that providing evidence regarding the merger paradox for Germany is a contribution to fill an existing gap – but is from a methodological point of view nothing new. So what makes my analysis special?

17 Fremdling and Krengel (1985) provided an excellent and critical overview on the importance of cartels in the German industry. They stated that the role of cartels is overstated regarding the impact on productivity, growth and price setting. They focused, however, mainly on the iron and steal industry.

18 Gömmel (1985) argued that more than three fourth of the newspapers criticized the formation of cartels and syndicates. Especially, the leading newspapers measured by circulation like the `Morgenpost´ in Berlin attacked heavily the collusive behavior.

19 Wengenroth (1985) described the general development of cartels in the German industry. A detailed discussion with regard to debates in parliament on collusive behavior was provided by Blaich (1973).

20 Banerjee, Eckhard (2001), Leeth, Borg (1994, 2000), and Borg et al. (1989)

2.2.2 What makes my analysis special?

In contrast to all former event-studies for the pre-World-War I period, I apply an event-study method based on daily returns, which is in line with studies covering later periods. Borg et al.

(1989) used monthly stock returns as well as Leeth and Borg (1994, 2000); however, Banerjee and Eckhard (2001) worked with weekly data – but this is still inferior compared to my precision in measuring market responses. Note that using monthly instead of daily returns makes it more difficult to detect abnormal stock price movements. Generally, the longer the chosen sampling interval the more cross-sectional units are needed to maintain a high power of the event-study. Morse (1984) presented a precise analysis on the usage of daily versus weekly respectively monthly returns.

Because I work on a daily frequency of my data, I have to determine the exact announcement day of a merger. This requires to read daily newspapers for a specific period of time – obviously, a time consuming task. Of course, daily newspapers are much more precise than weekly or monthly information sources that mainly focus on larger acquisitions.

Correspondingly, when one wants to analyze the success of mergers among smaller companies, this is only possible by reading daily newspapers. Despite the fascinating stories on mergers spread by daily newspapers, using this historical sources comes with a cost. I have to restrict my analysis to a predefined time interval to be able to read the newspapers during this period. In a subsequent section, I will mention the pros and cons of this method of sampling.

Besides accumulating new information regarding smaller transactions, which indisputably is an interesting contribution to the strand of literature in economic history, I also develop a new econometric tool for event-studies. This tool, a simultaneous equation approach, controls for the influence of the choice of the estimation period on the inference of factors that lead to more successful mergers. Applying this new idea to other event-studies seems to be worthwhile.

2.2.3 The event-study method: A brief review

Since its introduction²¹ into the field of empirical finance, the event-study method developed to one of the most often applied device to measure the economic impact of remarkable events.

This many-sided method was used for a growing number of applications; moreover, Binder

21 Especially Fama et al. (1969) and the study by Ball and Brown (1968) should be mentioned as pioneering applications of event-study methods.

(1985), Boehmer et al. (1991), Malatesta (1986), and Sefcik and Thompson (1986) developed some modifications of the basic event-study method.²²

Economic theory offers explanations how economic events should influence the firms' market value. The aim of event-studies is to measure this impact by comparing the stock price movement in the presence of events with the normally expected price development. To achieve an accurate measurement, market prices should fully reflect currently available information. This strong informational efficiency of the market²³ is a crucial prerequisite;

however, one can further relax this assumption by allowing an adaptation process of stock prices due to new information. This means that I have to define an event window during which the adaptation process should be finished.

2.2.4 The structure of this chapter

The remainder of this chapter is organized as follows. First, I thoroughly discuss the method of sampling; thereby, the pros and cons of my procedure are stressed. To give an impression regarding my data set, I present several descriptive figures and a selected number of brief case studies. Thereafter, I introduce the theoretical background of my event-study approach;

thereby, I point out to what extent basic models should be modified to cope with special historical shortcomings. After making these modifications, I estimate the normal returns on the basis of the chosen estimation window by applying the constant-mean-return model (CMR).²⁴ Finally, I calculate the abnormal returns occurring in the event period and assess their significance. In a cross-sectional model, I try to figure out which explanatory variables influence the success of mergers and close with a brief discussion of my results.

2.3 The method of sampling

2.3.1 What can be regarded as a merger?

There are different kinds of transactions that could be called a merger, for instance subsidiary mergers or consolidations – but this investigation only deals with mergers after which the acquiring company survive in a legal manner, whereas the target firm becomes a part of the acquirer. The task to distinguish between a merger that fulfills this requirement and other forms of collusive behavior is sometimes tricky. During the pre-World-War I period and afterwards, several forms of non-tacit collusion existed,²⁵ Tilly (1982) stressed

22 Armitage (1995) provided an excellent overview regarding the modifications and often used basic models.

23 See Fama (1970), p. 383.

24 See Masulis (1980).

25 See also Feldenkirchen (1988).

the importance of pooling agreements (`Interessengemeinschaften´) that cannot be regarded as mergers. Nevertheless, such pooling agreements can be the starting point for a further integration of companies and, consequently, can lead to an actual merger. Besides being the potentially first step toward full unification, `pooled´ companies can together acquire a competitor. As an illustration, table 2.1 summarizes the newspaper articles that dealt with the merger of `Höchst´ and `Kalle & Co. AG´; thereby, the pooling agreement with

`Leopold Casella & Co´ played a crucial role in financing the acquisition.

Table 2.1: The case of `Höchst´ and `Kalle & Co. AG´

Date of the newspaper announcements

Newspaper announcements in chronological order

`Berliner Börsenzeitung´

Information provided by the year book `Handbuch der deutschen

Aktiengesellschaften´

11^th April 1908 Morning issue, insert II

“`Höchster Farbwerke´ acquires

`Kalle & Co. AG´ by issuing new shares. Already prior to this announcement rumors spread about an impending increase in nominal capital of `Höchster Farbwerke´”

12^th April 1908 Sunday issue, page 15

“`Höchster Farbwerke´

announces its annual accounts and also stresses that the management of the target firm should stay in charge after the acquisition of `Kalle & Co. AG´”

13^th April 1908 Morning issue, insert IV

“`Höchster Farbwerke´ convenes a shareholder gathering to decide about the issue of new shares and the acquisition. The gathering will take place on 9^th May 1908”

19^th April 1908 Sunday issue, insert II

“The shareholder gathering of

`Kalle & Co. AG´ will be held on 11^th May 1908”

9^th May 1908 Evening issue, page four

“The shareholder gathering of

`Höchster Farbwerke´ approves the increase of nominal capital by 10.5 million Mark. The nominal capital now reaches 36 million Mark.”

11^th May 1908 Morning issue, page 14

“The shareholder gathering of

`Höchster Farbwerke´ accepts the acquisition of `Kalle & Co. AG´.

A part of the acquired shares that represent 4 million Mark in nominal capital will be passed on to `Leopold Cassella & Co.

GmbH´ with which a pooling agreement exists”

12^th May 1908 Evening issue, page 11

“Shareholder gathering of `Kalle

& Co. AG´ approves the offer”

“To deepen the relation between

`Höchster Farbwerke´ and `Kalle

& Co. AG´ an agreement was signed in 1908 to acquire shares from former principal shareholders. This acquisition was undertaken together with

`Leopold Cassella & Co. GmbH´

and reached a volume of 4,000,000 Mark (nominal capital). The `Höchster Farbwerke´ now own shares with a nominal capital of 3,200,000 Mark, whereas `Leopold Cassella

& Co. GmbH´ own 800,000 Mark in nominal capital”

Source: The indicated issues of the `Berliner Börsenzeitung´ and the `Handbuch der deutschen Aktiengesellschaften´, issue for the years 1913-1914, volume I, page 1600-1602.

2.3.2 What can we learn from this case study?

I combine the information provided in the daily newspaper `Berliner Börsenzeitung´ and the year book `Handbuch der deutschen Aktiengesellschaften´²⁶ to get as much detail as possible about the initiated mergers. By reading the daily newspaper carefully, it is possible to capture the whole process of a merger that starts with the initial announcement and ends with the approval of the shareholders. In spite of the work intensity,²⁷ the `Berliner Börsenzeitung´ offers an excellent up to date information regarding events during the pre- World-War I period. For instance, the major results of the shareholder gathering of

`Höchster Farbwerke´ that took place on 9^th May 1908 were reported in the evening issue of the same day. Having in mind the long process and the many hurdles a merger has to overcome nowadays, the extremely short time span between the announcement, the call for the shareholder gatherings, and the approvals is astonishing. This high speed of decisions is a common feature of all of my detected mergers. Moreover, hostile takeovers or the replacement of the management were highly unusual in the pre-World-War I period.

Typically, the newspaper announcements also contain information whether the management is allowed to stay in office or not. Generally, the quality and the high detail provided by the daily newspaper is remarkable. Accordingly, the daily information enables to precisely determine the announcement day, which is crucial for measuring the market response triggered by newly available information.

2.3.3 Drawing a sample

Caused by the time intensive and meticulous work inherent with reading daily newspapers, I had to restrict the time period. Accordingly, I included all mergers announced between 1^st January 1908 and 31^st June 1908 into my initial sample. Encouraged by Tilly’s (1982) statements about drawing samples in historical time periods, this method is a widely accepted procedure.²⁸ The advantage is that all events are considered regardless if a firm is listed on the stock exchange respectively is very small. As a first step, I collect all relevant information and observe 101 announcements. However, to conduct an event-study, share prices have to be observed. This prerequisite together with the requirement that a sufficient amount of trades occurred reduce the number of included companies dramatically. Thus, I

26 The `Handbook of the German Companies´ contains firm specific information on earnings, dividend payments as well as special activities, for instance, stock splits. Of course, the handbook does not contain any details about announcement days in daily newspapers.

27 Note that the `Berliner Börsenzeitung´ had in these years a morning and an evening issue every working day; this makes the reading very time consuming. Even worse, the newspaper was also issued on Sundays.

28 Tilly (1982) argued that choosing a specific time interval during which as much information is accumulated as possible is an appropriate method of sampling in economic history.

end up with forty-five cross sectional observations. Nevertheless, this sample size is in line with other event-studies, or simulation experiments executed by Brown and Warner (1980, 1985). Therefore, if one bases the analysis on daily returns, the econometric power of an event-study is very large.²⁹ Despite the high analytical quality, I should discuss the `historical power´ of my investigation, which is done in a subsequent section.

Moreover, it is crucial to determine precisely the day of the announcement to avoid false measurement; therefore, actuality and objectivity of the information source are important criterions. For the American market, the day of publication in the Wall Street Journal is usually used as event day. In Germany in the year 1908, the newspaper that satisfies these criterions best is the `Berliner Börsenzeitung´ because it possessed a great importance for investors and – thanks to telegram announcements – a high actuality. Therefore, the date of the first publication in the `Berliner Börsenzeitung´ is defined as the event day. Besides the determination of event days, more data are needed.

The `Berliner Börsenzeitung´ delivers stock prices of the Berlin stock exchange as well as other regional exchanges on a daily basis. This is an important improvement in comparison to using monthly or weekly stock returns,³⁰ which is too rough and leads to considerable methodical problems, for instance, cross correlation is more likely to occur.³¹

For the cross-sectional models additional information on stock characteristics like firm size is needed. A reliable source for company specific information is the `Handbuch der deutschen Aktiengesellschaften´. This year book also contains information on merger activities; however, only the year of an acquisition is usually reported.

After determining the event day, I turn to specify the event period that begins fifteen days before the announcement and ends fifteen days thereafter. In a later section, I justify this choice. Moreover, to estimate the normal returns, I collect fifty daily returns for each stock of my sample from the period January and February 1907. This estimation period is far enough away from the merger announcement and, hence, is not affected by the events, which is a prerequisite for estimating normal returns. Note that the estimated normal returns reflect the stock price movement without the merger event.

29 See Morse (1984).

30 See, for example, `Berliner Börsen-Courier´ and `Neuman’s Kurs-Tabellen der Berliner Fonds-Börse´ that provide monthly data.

31 Bernard (1987) found this result by running simulation experiments.

2.3.4 Why should I choose the year 1908 as investigation period?

Despite the high empirical power of my event-study, economic historians may wonder why I choose the year 1908 for my analysis. Henceforth, this decision needs a justification. Because only short time intervals can be analyzed, one should choose an interesting period for discussing the merger paradox. Note that my aim is to test whether the merger paradox exists and, accordingly, whether acquiring companies gain from mergers. Consequently, my null hypothesis states that the merger paradox can be observed. The rejection of the null hypothesis is more difficult in periods in which risky investments and, hence, mergers are punished by the market. If the market is bearish, such a punishment seems to be more likely.

Besides this argument, one can also point to the fact that in the year 1907, which serves as estimation period, higher normal returns should result.³² Accordingly, in the downturn of the market in 1908, it should be more difficult to observe positive abnormal returns and, thus, to reject the `merger paradox´. Figure 2.1 depict the `Donnerindex´ to illustrate the basic trend in the market.³³ Although this index is often criticized for its composition, it should just give a first impression regarding the general market situation. In chapter five, I will construct my own market index and can overcome typical pitfalls of the

`Donnerindex´.

In addition, working with daily returns typically yields estimated normal returns that are not significantly different from zero. The following section provides my empirical findings for the normal returns and corresponding confidence intervals. In addition, chapter four deepens the discussion further and tests whether changes regarding the length or the location of my estimation period matter.

32 Note that I used January and February 1907 as estimation period during which the `Donnerindex´ reached values between 122.36 and 121.03. Thereafter, a pronounced decline set in, and the market lost more than 10%

till the beginning of my event period; however, this decrease was not included in my estimation period.

Furthermore, my fourth chapter provides evidence that switching the estimation period (using March and April 1907) or extending the period does not affect my results.

33 For instance, Grabas (1992) listed the `Donnerindex´ for this period in her data appendix; thus, I use this information to depict the market index in figure 2.1.

Figure 2.1: The `Donnerindex´ on a monthly basis 1895 to 1913

The event-period, 1^st January to 30^th June 1908, reaches a relatively low share price level compared to the previous estimation period, January to February 1907. On average, the estimation period exhibits 10% higher share prices in comparison to the event-period.

Donner index

80 85 90 95 100 105 110 115 120 125 130

189 5

189 6

189 7

189 8

189 9

190 0

190 1

190 2

190 3

190 4

190 5

190 6

190 7

190 8

190 9

191 0

191 1

191 2

191 3

2.3.5 Descriptive analysis

To obtain a first impression of my data set, I divide the observed merger announcements into several subgroups according to line of business, success of merger, whether the management of the target firm changed after the takeover, how the deal is financed, the number of involved firms, and the motives behind the decision for initiating a merger, if this is published. Caused by missing values, I can only include 79 out of 101 cases in my descriptive analysis.

During my investigation period, I detect 101 mergers; however, the merger activity is far from being stable over the six considered months. Figure 2.2 shows that the peak is reached in April 1908. During this month 29.70% of all mergers are executed. Thus, I can conclude that the merger activity is time varying.

Figure 2.2: Time-varying merger activity over the six included month

Figure 2.2 depicts the number of mergers released during the respective week; thereby, the study starts in the first week of January 1908 and ends with the last week of June 1908.

0 1 2 3 4 5 6 7 8 9 10

1 3 5 7 9 11 13 15 17 19 21 23 25 27

week

number of cases

Table 2.2 uncovers discrepancies in merger activity that depend on the affiliation to a specific line of business. I make a crude distinction among the major groups in an economy, namely service, manufacturing, and raw material production. This is further refined; thereby, the subgroups are the largest groups in the respective sector. It is apparent that especially the banking industry is very active in undertaking mergers, whereas other industries like the mining sector exhibit only weak activities. This might be due to the formation of syndicates in the mining industry; accordingly, a subsequent section highlights this issue.

Table 2.2: Merger activity in different lines of business

Table 2.2 summarizes the number of mergers within each indicated line of business and the contribution, in per cent, of the respective category to the whole merger activities.

Line of Business Absolute number of mergers In per cent of total mergers

Banking 31 39.2%

Real estate 3 3.8%

Traffic 8 10.1%

Mergers in service industry 42 53.2%

Chemical 2 2.5%

Electrical engineering 3 3.8%

Brewery 4 5.1%

Other manufacturing industries 20 25.3%

Mergers in manufacturing 29 36.7%

Coal 3 3.8%

Potash 2 2.5%

Other raw material production 3 3.8%

Mergers in raw material production 8 10.1%

Total number of mergers 79 100%

In addition, I can determine the number of firms that are involved in a merger. In almost all cases only two firms interact; nevertheless, there are five out of 79 cases in which more than two firms merge. Due to the requirement that a company must be listed on a German stock exchange to observe daily returns, the number of observations drops to fifty. These companies are included in my event-study. However, lacking information on the details of the transaction not provided by the daily newspaper forces me to reduce the number of observations to forty- five for my cross-sectional models. Note that the way of financing a merger stays in five cases unfortunately unclear. Nevertheless, the merger paradox can be discussed by using the figures based on the event-study or by estimating the partial impact of being an acquirer in the cross- sectional analysis. Therefore, my analysis is by no means limited. Because the forty-five companies used in my cross-sectional model are of primary interest for my investigation, table 2.3 presents some descriptive figures. These figures include the line of business, whether the company is an acquirer or target, how the deal is financed, the decision of the shareholders and whether the management is replaced. Based on annual information provided by the

`Handbuch der deutschen Aktiengesellschaften´, table 2.3 shows the market capitalization in

million Mark. The thirteen target firms possessed an average market capitalization of 16.47 million Mark, whereas the 32 acquirers reached on average 46.45 million Mark. However, I cannot state that acquirers are about three times larger than their `victims´ in a transaction because not all targets are listed on the German stock exchanges.

Most announced mergers are successful; thereby, success means that the merger is executed after its public declaration. The advisory board and an extraordinary shareholder gathering must agree to the proposed merger. The minimum proportion of the shareholders that have to vote for the merger is defined by the statues of the specific company. Only two mergers fail to achieve the necessary majority.

It was uncommon to replace the management of the target firm after the merger, although the replacement of an inefficient management is often seen as one source of efficiency gains from mergers. Because only three cases can be observed in which the management is fired, the existence of a market for corporate control³⁴ and corresponding incentives for managers can hardly be supported by my data.

About 55.7% of the mergers were financed by cash payments, and the rest was conducted by using own shares as `acquisition currency´. This crude distinction seems to be a little bit misleading because cash payments are sometimes accompanied by an offer to transfer shares. Therefore, I regard a merger as being financed by cash payment if cash is the predominant payment – more than 90% of the total offer. Moreover, I observe that smaller acquisitions are more likely to be financed by cash payments, and if the target is not a listed company, cash payment is common.

2.3.6 The role of cartels and syndicates

I should stress that the formation of cartels and syndicates is not taken into account because they do not act like one firm after their formation. Despite this fact, cartels and syndicates play an important role for deciding to undergo a merger. Collusive behavior should reduce competition; therefore, one main motive to merge is lacking in industries in which cartels and syndicates prevail. Before suggesting that cartels reduce the driving force to merger, I should consider the spirit industry in which a syndicate called `Spirituszentrale´ existed. This syndicate possessed a strong market position especially in the southern regions of Germany.

The `Nürnberger Presshefen- und Spiritusfabriken AG´ decided on 19^th April 1908 to leave the syndicate. Because the management was afraid of retaliations, they bought two additional plants, one near Berlin, the other in Breslau, to increase firm size and to serve the local

34 See Manne (1965).

markets in Northern Germany. According to this case study, the impact of cartels and syndicates cannot be determined in a clear manner.

Besides this `narrative´ evidence that the effect of cartels is unclear, econometric concerns make an evaluation of the partial impact less convincing. As far as I know, an unique measure for the degree of collusion that is applicable for every industry does not exist.

Arguing in terms of average production cost and output prices achieved by a syndicate may work for the iron and steal industry³⁵ – but cannot be applied to the banking industry.

Moreover, also a practical concern arises because a considerable number of companies included in my cross-sectional models belong either to the banking or to the mining industry.

35 Krengel (1982) provided several measures for the iron and steal industry regarding the degree of concentration and collusive behavior.